The Social Relationship Context of Elder Mistreatment

Abstract

Background and Objectives

Elder mistreatment victims at risk of poor physical and psychological health may benefit from increased social support. This article identifies mistreatment victims among community-dwelling older Americans and maps their social networks to guide the design of social support interventions.

Research Design and Methods

Using nationally representative survey data from Wave 3 (2015–2016) of the National Social Life, Health, and Aging Project (N = 2,334) and descriptive, latent class, and regression analyses, we estimate the prevalence of mistreatment since age 60, identify the alleged perpetrators’ relationships to the victims, and examine victims’ social networks.

Results

Self-reported lifetime prevalence of elder mistreatment is as high as 21%, depending on the mistreatment behavior measured. Latent class analysis reveals two mistreatment classes: 12% of older adults experienced multiple types of mistreatment (polyvictimization), and 6% experienced primarily financial mistreatment. Although alleged perpetrators are unlikely to appear in older adults’ core social networks, the most commonly reported perpetrators are children and relatives. Regression analyses show that experiencing mistreatment since age 60 is associated with having less current social support, more social strain, and fewer kin in the core social network. Older adults reporting polyvictimization also have less-dense core networks.

Discussion and Implications

Increasing family support should be done cautiously because children and relatives are frequently named as mistreatment perpetrators. Increasing communication across polyvictimization victims’ network members may support their well-being. Providing outside assistance with financial management could benefit financial mistreatment victims.

Elder mistreatment victims are at risk of poor psychological and physical health, including loneliness (Wong & Waite, 2017), depression (Luo & Waite, 2011), anxiety (Luo & Waite, 2011; Wong & Waite, 2017), poor self-rated health (Amstadter et al., 2010), disability (Schofield, Powers, & Loxton, 2013; Wong & Waite, 2017), and mortality (Dong et al., 2011; Schofield et al., 2013). Consequently, some scholars advocate for increasing victims’ social support (Acierno et al., 2010), as social relationships can promote health (Thoits, 2011), buffer the ill effects of mistreatment (Dong et al., 2011; Luo & Waite, 2011), and reduce the risk of further mistreatment (Acierno et al., 2010; Cisler, Begle, Amstadter, & Acierno, 2012; Schafer & Koltai, 2015).

Yet, we cannot make specific recommendations for increasing social support without first knowing the quality (i.e., social support and strain) and structure (i.e., size, density, and composition) of mistreatment victims’ broader social relationship and social network contexts. First, we need to confirm that victims’ social relationships differ qualitatively from that of nonvictims. If victims perceive lower support—and perhaps also higher strain, as strain is conceptually distinct from support (Chen & Feeley, 2013) and could have implications for designing relationship-centered interventions—then we can examine victims’ social network structure to provide targeted intervention suggestions. For instance, although family members are often key members of older adults’ core social networks (i.e., those with whom they discuss important matters; Cornwell, Laumann, & Schumm, 2008), recommending that victims turn to family members for support may be unhelpful if they are sources of strain or abuse (Schafer & Koltai, 2015). To design effective social support interventions, we need to understand more about elder mistreatment victims’ social context.

We advance hypotheses about victims’ social networks and social relationships based on socioemotional selectivity theory (Carstensen, Fung, & Charles, 2003) and previous elder mistreatment theory and research. Socioemotional selectivity theory suggests that as individuals age, they trim unsupportive or conflictual ties from their social networks and maintain their more supportive ones. This is because, “As people age, they realize that time is ‘running out,’ and begin to focus on the present as opposed to the future” (p. 107). Thus, although removing negative ties may alter the structure of a network by shrinking it, older adults’ remaining relationships may be qualitatively more supportive and meaningful. Given that perpetrator–victim relationships can be strained or conflictual, socioemotional selectivity theory suggests that elder mistreatment victims may remove perpetrators from their networks.

However, elder mistreatment research suggests that many perpetrators are “trusted others” (Bonnie & Wallace, 2003) who may be family members caring for older adults (Gordon & Brill, 2001). Such perpetrators can include adult children who are dependent on or living with older parents (DeLiema, Yonashiro-Cho, Gassoumis, Yon, & Conrad, 2018; Lachs & Pillemer, 2004; Schafer & Koltai, 2015). Such obligatory family ties may be difficult to completely remove from a social network (Offer & Fischer, 2018). Therefore, we predict that family caregivers and coresident adult children may be prominent actors in mistreatment victims’ social lives, but we test whether (rather than assume that) they provide social support.

The inability to remove a perpetrator from a network may also affect the quality and structure of other relationships beyond the perpetrator–victim dyad. Others connected to the victim may be uncomfortable talking to the perpetrator, thus decreasing network density and dampening flows of trust across the network. Some research indicates that victims can have structurally and qualitatively weaker social networks than those without mistreatment experience. Structurally, victims are less likely to have a spouse or partner (Laumann, Leitsch, & Waite, 2008) and are embedded in less-dense networks (Schafer & Koltai, 2015). Qualitatively, victims report lower support from their social ties (Acierno et al., 2010; Beach, Schulz, & Sneed, 2018). Thus, we hypothesize that victims’ social networks may provide limited psychosocial resources, which poses a challenge to designing effective interventions.

This study examines the social network and relationship context of elder mistreatment victims in a nationally representative sample of older Americans to guide the development of more specific social support interventions for those with mistreatment experience. We use the unique social network and elder mistreatment data from Wave 3 of the National Social Life, Health, and Aging Project (NSHAP) to accomplish three objectives. First, we estimate the proportion of older adults with mistreatment history. We distinguish among different types of mistreatment, as victims who experienced particular patterns of mistreatment may have different social contexts and different social support needs. Second, we describe the reported perpetrator–victim relationships to understand how perpetrators fit into older adults’ social networks. Finally, we characterize the quality and structure of victims’ wider social relationship and social network contexts to pinpoint promising ways to promote social support.

Methods

Data and Sample

NSHAP began as a nationally representative longitudinal study of community-dwelling older adults in the United States born between 1920 and 1947 (Waite et al., 2017; Waite, Cagney, et al., 2014; Waite, Laumann, Levinson, Lindau, & O’Muircheartaigh, 2014). Collaborating with the Health and Retirement Study (HRS) in 2004, NSHAP used multistage area probability sampling to select respondents. Men, those aged 75–84, African Americans, and Hispanics were oversampled. The original NSHAP cohort was surveyed for Wave 1 in 2005–2006 (N = 3,005). In Wave 2 (2010–2011), surviving respondents were re-interviewed, and a sample of spouses and coresident partners of Wave 1 cohort members, regardless of age, were also invited to participate (N = 3,377). In 2015–2016, 2,409 returning Wave 1 and 2 respondents and their partners were interviewed for Wave 3 (Cohort 1). Wave 3 also included 2,368 new respondents born during the Baby Boom and their partners (Cohort 2). In all waves, interviewers from NORC at the University of Chicago conducted in-home interviews with the older adult respondents. Participants were also asked to complete and return a self-administered leave-behind questionnaire. Overall response rates for the three waves were 75.5%, 74% and 71%, respectively. Sampling weights constructed by the NSHAP team (O’Muircheartaigh, Eckman, & Smith, 2009) account for both differential nonresponse and oversampling.

Our analytic sample included 2,334 Cohort 1 Wave 3 respondents aged 60–95 with elder mistreatment history data. We excluded from analyses: 37 Cohort 1 Wave 3 respondents younger than 60 who were ineligible for the elder mistreatment questions; 38 Cohort 1 respondents without sampling weight information; and all Cohort 2 respondents because they did not receive the elder mistreatment module. Weighted descriptive statistics for our sample appear in Table 1.

Table 1.

Weighted Descriptive Statistics

Measure	Mean/%	Measure	Mean/%
Age (60–95)	75	Physical health	2.3
Gender (women)	57%	(Poor/fair = 1 to excellent = 4)
Race/ethnicity		Has spouse or partner	68.9%
White	82%	Quality of social relationships
Black	9%	Spouse support (1–4)	3.7
Hispanic, non-Black	7%	Family support (1–4)	3.4
Other	2%	Friend support (1–4)	3.0
Education		Spouse strain (1–4)	2.1
Less than high school	14%	Family strain (1–4)	1.9
High school	25%	Friend strain (1–4)	1.6
Vocational/some college	33%	Structure of social relationships
Bachelors or more	29%	Core social network size (0–5)	3.9
Functional status		Core social network density (0–1)	0.8
No limitations	48%	% Kin in core social network	64%
IADLs only, no ADLs	31%	% Children in core social network	31%
Any ADLs	21%	Lives with a child	14%
Cognition (MoCA-SA, 0–20)	14.1	N	2,334

Notes: ADLs = Activities of Daily Living; IADLs = Instrumental Activities of Daily Living; MOCA-SA = Survey-Adapted Montreal Cognitive Assessment (Kotwal et al., 2015). Spouse support and strain measures apply only to partnered respondents.

Independent Variables—Elder Mistreatment History

The elder mistreatment module in NSHAP Wave 3 consisted of 10 main questions as well as a series of follow-up questions about older adults’ mistreatment experiences. The 10 questions were identical to those designed for the Population Study of ChINese Elderly in Chicago (PINE) Screener (Dong, Wong, & Simon, 2014). The PINE Screener adapted questions from the Hwalek–Sengstock Elder Abuse Screening Test (H-S/EAST; Hwalek & Sengstock, 1986) and the Vulnerability to Abuse Screening Scale (VASS; Schofield & Mishra, 2003). It had good reliability (α = .80) and cross-cultural applicability (Dong, Chen, Fulmer, & Simon, 2014).

The mistreatment module opened with: “For this next section, please think about ways that people behave toward you that bother you. Specifically, think of people and your relationships with them. Since you turned 60…”

1. Have you felt uncomfortable with anyone in your family?

2. Have you felt that nobody wanted you around?

3. Has anyone told you that you gave them too much trouble?

4. Have you been afraid of anyone in your family?

5. Has anyone close to you tried to hurt or harm you?

6. Has someone in your family made you stay in bed or told you that you are sick when you know you are not?

7. Has anyone close to you called you names or put you down or made you feel badly?

8. Has anyone forced you to do things you didn't want to do?

9. Has anyone taken things that belong to you without your OK?

10. Has anyone borrowed your money without paying you back?

Following the PINE Screener, NSHAP Wave 3 used the “since turning 60” approach to characterize “lifetime prevalence” (Beach, Schulz, Castle, & Rosen, 2010) or history of elder mistreatment. As such, our results tell us about the association between mistreatment since age 60 and current social network characteristics to guide our intervention suggestions for those with a history of mistreatment.

We coded answers No = 0, Yes = 1, and “Don't Know” and “Refused” as missing for each of these questions. Missingness on these measures ranged from 0.04% (N = 1; Has someone in your family made you stay in bed or told you that you are sick when you know you are not?) to 0.64% (N = 15; Have you felt that nobody wanted you around?). Because answering “yes” to a single question could overidentify older adults as mistreatment victims—a limitation of these broad questions—we conducted robustness checks classifying mistreatment according to seriousness based on a follow-up question assessing mistreatment severity (Burnes, Pillemer, & Lachs, 2016). Major findings remained qualitatively the same (analyses available on request). Because elder mistreatment is usually under-reported rather than over-reported (Acierno et al., 2010), we present results based on analyses of all mistreatment experiences to increase statistical power. We discuss whether these robustness checks produced similar or different results from our main analyses throughout the results section.

Answering “yes” to a mistreatment question prompted respondents to identify the relationship of the alleged perpetrator to the respondent. First, to determine whether the perpetrator was a current member of the victim’s core social network (someone with whom they discuss important matters), elder abuse victims were asked, “Thinking about the person who has done this the most since you turned 60, is this person someone we wrote down on your roster earlier?” This “roster” was the social network roster completed at the beginning of the NSHAP interview (Cornwell, Schumm, Laumann, & Graber, 2009). To generate the roster, respondents were first asked, “From time to time, most people discuss things that are important to them with others. For example, these may include good or bad things that happen to you, problems you are having, or important concerns you may have. Looking back over the last 12 months, who are the people with whom you most often discussed things that were important to you?” This question identified respondents’ current core social network members—up to five people. To identify other potentially important people who may regularly interact with these older adults, respondents were asked to name their spouse or romantic partner if they had not already done so and then were asked to add any household members who they had not already named to their roster. We refer to these additional roster members as respondents’ household members. All others who may be connected to the individual but were not included on the roster—other family members, coworkers, neighbors, or members of organized groups to which respondents belong, for example—we call others.

For elder mistreatment victims who indicated that the perpetrator was a core social network member or a household member, we used linked relationship data to characterize the perpetrator–victim relationship. For victims who indicated the perpetrator was neither a core social network member nor household member (i.e., an other), respondents were asked to identify the relationship of the perpetrator to them based on a list of 19 kin and nonkin relationships and an open-ended option. To facilitate analysis, we condensed the perpetrators’ identities, regardless of whether they were core social network members, household members, or others, into four broader categories based on previous literature (Acierno, 2003; Laumann et al., 2008; Lindau, Laumann, Levinson, & Waite, 2003): (i) current spouses or partners; (ii) children and step-children; (iii) other relatives including parents, siblings, all in-laws, and former family members; and (iv) nonkin perpetrators such as friends, neighbors, bosses, therapists, and health care providers. Three elder mistreatment questions specifically asked about mistreatment by family members (items 1, 4, and 6), precluding the perpetrator from being classified as a nonkin perpetrator. All results were robust to the exclusion of these three items (results available on request), so we retained them in all analyses to maintain the 10-question elder mistreatment module as a whole.

Dependent Variables—Social Context

To confirm that victims lacked support and experienced more strain, we estimated older adults’ levels of social support and strain from spouses, family members, and friends. Social support and strain questions referred to spouses, family members, and friends in general, not just to individuals on respondents’ network rosters, allowing us to gauge victims’ broader social context. The social support scales were created for this study using two items: “How often do you open up to ___ if you need to talk about your worries?” and “How often do you rely on ___ for help if you have a problem?” They were asked separately for spouses, family members, and friends. Responses ranged from Never = 1, Hardly ever or rarely = 2, Some of the time = 3, to Often = 4. The two items were averaged into 4-point social support scales for spouses (α = .63), family members (α = .73), and friends (α = .74). The social strain measures for spouses (α = .66), family members (α = .56), and friends (α = .60) were similarly constructed by averaging the following two items: “How often does ___ make too many demands on you?” and “How often does ___ criticize you?” Higher scores reflected higher support and higher strain, respectively. Those without spouses or partners were excluded from the measures and analyses of spousal support and strain.

To examine the structural characteristics of victims’ social context to guide specific intervention recommendations, we first estimated older adults’ core social network size. Then, we calculated the density (Cornwell et al., 2009) of older adults’ core social networks. Density measures whether core social network members had ever spoken to each other and was reported as the proportion of all possible alter–alter ties that existed in the core social network. Next, we estimated the percentage of older adults’ core social network members who were kin, and the percentage who were children. Finally, we reported the proportion of older adults with coresident children. Core social network members who were children factored into calculations for percentage of core social network members who were kin, as well as calculations for percentage of core social network members who were children. The proportion of older adults with coresident children may or may not have included children who were also core social network members; it was possible to live with a child (i.e., a household member), but not consider that child a core social network member.

Covariates

Regression models controlled for demographic and health characteristics associated with both elder mistreatment and social context (Lachs & Pillemer, 2004; Laumann et al., 2008): age, gender, race/ethnicity, education, marital/partnership status, self-rated physical health, functional status, and cognition. Table 1 shows the operationalization of all key measures.

Analysis

First, we calculated survey-weighted proportions of older adults who answered “yes” to each of the 10 elder mistreatment questions. Next, we identified whether the perpetrator of each mistreatment behavior was a core social network member, a household member, or an other. Then, we specified whether perpetrators were spouses, children, other family members, or nonkin, regardless of network membership status.

Next, we used the mistreatment questions to identify latent classes (McCutcheon, 1987) of elder mistreatment to understand whether social relationship contexts varied by pattern of mistreatment history. One-class, two-class, and three-class models with a logit link were compared using Akaike’s information criteria (AIC), where smaller AIC values were better. We chose the model with the smallest AIC value, estimated the posterior probability of class membership for each respondent, and then assigned them to the class for which the posterior probability was >.50. In cases where all probabilities were <.50, class membership was assigned based on the highest probability. For example, in a three-class model where a respondent had posterior probabilities of .31 for being in class 1, .45 for being in class 2, and .24 for being in class 3, they were assigned to class 2.

We then used this mistreatment class measure in a series of regressions to predict differences across the social relationship context measures. We used OLS regression for (i) spousal support; (ii) family support; (iii) friend support; (iv) spousal strain; (v) family strain; and (vi) friend strain to assess differences in social support and strain across mistreatment classes. Then, we examined differences across mistreatment classes in the structure of social relationships and networks. We used OLS regression to predict: (i) core social network size; (ii) core social network density; (iii) the percentage of older adults’ core social network that were kin; and (iv) the percentage of older adults’ core social network that were children. We used logistic regression to predict the odds of living with an adult child. All regressions controlled for the covariates described above except the models predicting spousal support and strain, which were restricted to partnered respondents and did not control for marital/partnership status.

Sample sizes varied across regressions due to such model-specific restrictions: for example, regressions predicting network density excluded those with zero or one alter in their core social networks because density calculations required at least two alters to exist in a network. Additional cases were lost from the analytic sample due to missingness in the functional status variable. Robustness checks excluding this measure from analyses produced similar results, so we kept this control in the final models. All descriptive, latent class, and regression analyses were then repeated with the more stringent measure of serious mistreatment to check the robustness of our findings (results available on request). Analyses were conducted in Stata 15 (StataCorp, 2017) using survey weights to account for differential probabilities of selection and nonresponse and to generalize to the community-dwelling population of older Americans aged 60–95 in 2015–2016.

Results

Prevalence and Perpetrators of Elder Mistreatment

Figure 1 depicts proportions of older adults who experienced specific mistreatment behaviors since turning 60. The height of each bar represents the proportion of older adults who answered affirmatively to each of the 10 mistreatment questions. Twenty-one percent of older adults reported having felt uncomfortable with someone in their family since they turned 60. Twenty-one percent also reported that, since turning 60, someone borrowed their money without paying them back. Other common mistreatment experiences included being called names, being put down, or having been made to feel badly (13%), and having belongings taken without permission (11%).

Figure 1.

Mistreatment by Core Social Network and Household Members.

Each bar is subdivided to show the proportion of perpetrators that were core social network members, household members, and others. Most perpetrators were neither core social network members nor household members. For example, 21% of those who made older adults feel uncomfortable were older adults’ core social network or household members, but the other 79% were neither. One exception to this pattern appeared for the measure, “Has someone in your family made you stay in bed or told you that you are sick when you know you are not?” Eighty percent of perpetrators were core social network or household members.

Figure 2 identifies perpetrators in more detail. Children and other relatives (regardless of their network membership status) were the most common perpetrators across most mistreatment behaviors, even when excluding items 1, 4, and 6 that ask specifically about mistreatment by family members. For example, when asked to identify who borrowed older adults’ money without paying it back, 54% indicated that their children were the perpetrators and 29% reported that other family members were the perpetrators. Nonkin were seldom reported as perpetrators, although 20% of those who took older adults’ things and 18% of those who borrowed older adults’ money without repaying it were nonkin. These descriptive findings identifying the most common mistreatment behaviors and perpetrators remained qualitatively the same in the analyses of serious mistreatment reports.

Figure 2.

Identity of mistreatment perpetrators.

The Broader Social Relationship Context of Elder Mistreatment Victims

To map the broader social network and relationship context of elder mistreatment, we first used latent class analysis on the 10 mistreatment questions to identify a typology of mistreatment victims. We found three groups of older adults distinguished by mistreatment experience since age 60. The largest class, 82% of older adults, had low likelihoods of experiencing any mistreatment since turning age 60. We call this class “mistreatment unlikely.” A second class representing 12% of older adults had higher odds of experiencing nearly all 10 mistreatment behaviors since turning 60. We call this the “polyvictimization” class because they experienced emotional mistreatment (e.g., 64% reported that someone close to them called them names or put them down or made them feel badly and 51% felt that nobody wanted them around), financial mistreatment (41% had money borrowed without it being repaid), and physical mistreatment (15% said someone close to them tried to hurt or harm them). The final class, representing 6% of older adults, we call the “financial mistreatment” class because of their higher likelihood of experiencing only financial mistreatment. Supplementary Appendix A lists the proportion of older adults in each mistreatment class who reported experiencing each of the 10 individual mistreatment behaviors. An additional latent class analysis of serious mistreatment produced the same three classes of elder mistreatment victims with 88% agreement between membership in the three classes based on all mistreatment reports compared with membership in the three classes based on serious reports only.

Regressions confirmed that the quality and structure of older adults’ social relationships and networks differed by mistreatment history. Results in Table 2 show that both mistreated classes reported less spouse and family support and more spouse and family strain than the mistreatment unlikely class. For example, the polyvictimization class scored 0.18 points lower on the spousal support scale (Model 1, p < .01), 0.25 points lower on the family support scale (Model 2, p < .001), 0.44 points higher on the spousal strain scale (Model 4, p < .001), and 0.53 points higher on the family strain scale (Model 5, p < .001) than the mistreatment unlikely class. Robustness checks using the serious mistreatment latent class measure produced similar findings, though the negative association between polyvictimization and spousal support did not reach statistical significance at the p < .05 level.

Table 2.

Quality of Social Relationships by Mistreatment Class

	Model 1:	Model 2:	Model 3:	Model 4:	Model 5:	Model 6:
	spousal support (1–4)	family support (1–4)	friend support (1–4)	spousal strain (1–4)	family strain (1–4)	friend strain (1–4)
Mistreatment class
Mistreatment unlikely (ref.)
Polyvictimization	−0.18**	−0.25***	−0.01	0.44***	0.53***	0.23***
Financial mistreatment	−0.21*	−0.28**	−0.25*	0.35***	0.34**	0.09
Age	−0.00	0.00	−0.01**	−0.00	−0.01**	−0.00
Gender
Men (ref.)
Women	−0.08*	0.32***	0.25***	−0.30***	0.03	−0.12***
Race/ethnicity
White (ref.)
Black	−0.15*	−0.22**	−0.12	0.22*	0.14	0.20**
Hispanic, non-Black	−0.06	−0.06	−0.23**	0.04	−0.00	0.07
Other	−0.00	−0.04	−0.26	0.13	0.29**	0.21*
Education
<High school (ref.)
High school	0.01	0.05	0.08	0.24*	0.16*	0.16**
Vocational/some college	0.11	−0.01	0.16*	0.26*	0.20**	0.21***
Bachelors or more	0.03	−0.05	0.09	0.27*	0.20*	0.19**
Physical health	0.03	0.07**	0.09***	−0.09**	−0.03	−0.01
Functional status
No limitations (ref.)
IADLs only, no ADLs	−0.10	−0.09	−0.10	0.01	0.08	−0.03
Any ADLs	−0.03	−0.08	0.03	0.03	0.10	0.03
Cognition	0.00	−0.01	0.00	0.01	0.01	0.01
Relationship status
Not partnered (ref.)
Partnered	—	−0.07	0.14**	—	−0.09	0.01
N	1,403	1,786	1,837	1,404	1,859	1,859

Notes: ADLs = Activities of Daily Living; IADLs = Instrumental Activities of Daily Living. Numbers in table are OLS regression coefficients. Spouse support and strain analyses apply only to partnered respondents.

*p < .05, **p < .01, ***p < .001.

Patterns for friend support and strain differed by type of mistreatment history. The polyvictimization class reported similar levels of friend support as those without mistreatment experience (Model 3, p > .05), but scored 0.23 points higher on the friend strain scale (Model 6, p < .001). The financial mistreatment class scored 0.25 points lower on the friend support scale than those without mistreatment experience (Model 3, p < .05; β = −0.15 and p = .057 in robustness checks), but was not statistically different from the mistreatment unlikely class in friend strain (Model 6, p > .05). Analyses using the serious mistreatment latent class measure produced the same pattern of findings and even showed a statistically significant association between financial mistreatment and increased friend strain.

Turning to the structural characteristics of older adults’ social networks in Table 3, we found that network size did not vary across the mistreatment classes. However, polyvictimization victims had less-dense core social networks (Model 2, β = −0.09, p < .01), and had fewer kin in their core social networks (Model 3, β = −5.43, p < .05; β = −5.2 and p = .07 in robustness checks) compared with those who were not mistreated. Financial mistreatment victims also had fewer kin in their core social networks (Model 3, β = −8.28, p < .05). Furthermore, although the financial mistreatment class had fewer children in their core social networks than others (Model 4, β = −8.62, p < .01), they had higher odds of living with a child (Model 5, odds ratio = 2.55, p < .001). Findings for the structural characteristics of mistreatment victims persisted in analyses restricting mistreatment to serious reports, although the positive association between financial mistreatment and living with a child did not reach statistical significance at the p < .05 level.

Table 3.

Structure of Social Relationships by Mistreatment Class

	Model 1:	Model 2:	Model 3:	Model 4:	Model 5:
	Core social network size (0–5)	Core social network density (0–1)	% kin in core social network (0–100)	% Children in core social network (0–100)	Lives with child (1 = Yes)
Mistreatment class
Mistreatment unlikely (ref.)
Polyvictimization	0.11	−0.09**	−5.43*	−3.86	0.81
Financial mistreatment	−0.01	−0.04	−8.28*	−8.62**	2.55***
Age	0.00	0.00	0.16	0.60***	0.99
Gender
Men (ref.)
Women	0.33***	−0.01	1.81	3.91*	0.87
Race/ethnicity
White (ref.)
Black	−0.47**	0.04	5.48	−0.77	1.31
Hispanic, non-Black	−0.22	0.02	11.16**	12.48**	2.90***
Other	0.03	0.05	0.96	2.70	3.82**
Education
<High school (ref.)
High school	0.19	0.01	4.46	7.42*	1.26
Vocational/some college	0.36**	−0.01	−3.11	2.33	1.04
Bachelors or more	0.44**	−0.04	−3.96	1.32	0.71
Physical health	0.09*	−0.00	−0.34	−0.14	0.91
Functional status
No limitations (ref.)
IADLs only, no ADLs	−0.08	0.03	1.54	0.87	1.18
Any ADLs	0.30**	0.00	−2.66	−1.33	1.35
Cognition	0.05***	−0.00	−0.40	−0.14	0.98
Relationship status
Not partnered (ref.)
Partnered	−0.16	−0.07*	−8.79***	3.10	1.74**
N	2,024	1,867	1,998	1,998	2,009

Notes: ADLs = Activities of Daily Living; IADLs = Instrumental Activities of Daily Living. Numbers in table are OLS regression coefficients except for Model 5 where values are odds ratios.

*p < .05, **p < .01, ***p < .001.

Discussion and Implications

Elder mistreatment is associated with poor physical and psychological health (Amstadter et al., 2010; Luo & Waite, 2011; Schofield et al., 2013; Wong & Waite, 2017). Increasing victims’ social support may improve their well-being and reduce their risk of further mistreatment (Acierno et al., 2010; Schafer & Koltai, 2015). We used nationally representative data from community-dwelling older Americans to paint a detailed picture of mistreatment victims’ social network and relationship context to guide specific recommendations for social support interventions.

Supporting our predictions (Bonnie & Wallace, 2003; Carstensen et al., 2003; DeLiema et al., 2018; Gordon & Brill, 2001; Schafer & Koltai, 2015), descriptive findings showed that although perpetrators were not usually victims’ core social network or household members at the time of data collection, they were commonly victims’ own children or other family members. Family perpetrators of mistreatment may not appear in victims’ core social networks or households because mistreated older adults may use socioemotional selectivity (Carstensen et al., 2003) strategies to protect themselves, such as removing these perpetrators from their core social networks and households, when possible. Anyone advising mistreatment victims to seek support from children or relatives should recognize that not all family members are appropriate social resources for older adults. These older adults may benefit more from relationships with friends or professionals. Yet, because perpetrators were not usually victims’ core social network members, encouraging mistreated older adults to turn to people that they trust—–whether or not they are family members—–can bolster victims’ social support.

Twelve percent of NSHAP Wave 3 older adults had a history of polyvictimization and 6% had a history of financial mistreatment. Consistent with prior studies (Acierno et al., 2010; Beach et al., 2018), mistreatment victims reported less social support and more strain than others. Perhaps those with lower support and higher strain had a greater risk of being mistreated in the first place (Beach et al., 2018; Luo & Waite, 2011). It is also possible that removing network members—but perhaps not completely (Offer & Fischer, 2018)—disrupted the network’s balance (Rawlings & Friedkin, 2017), hampering its ability to provide support and creating strained relationships among its members. The latter explanation complicates socioemotional selectivity theory because negative but obligatory ties may not be easily trimmed and removing conflictual ties may not always result in more supportive networks.

As both mistreated groups reported less support in their current spousal and family relationships than those without mistreatment history, interventions to increase support to victims should emphasize building trust and increasing reliability of nonperpetrator family members. Financial mistreatment victims in particular may benefit from interventions aiming to increase support from friends, as those with a history of financial mistreatment reported lower levels of friend support but polyvictimization victims did not.

Both types of mistreatment victims also reported higher social strain from spouses and family members. The polyvictimization class further reported higher strain from friends. Considering social strain along with social support suggests that efforts at increasing victims’ support alone may not be enough to bolster their health. Social support and social strain are independently associated with well-being (Chen & Feeley, 2013) and relationships characterized by ambivalence—both high support and high strain—are uniquely associated with worse health outcomes (Lee & Szinovacz, 2016). Social support interventions for mistreatment victims therefore need to consider ways to reduce social strain while improving support. Reducing friend strain among polyvictimization victims is particularly important.

Analyses of the structural characteristics of mistreatment victims’ relationships showed that network size did not differ by mistreatment experience. This encouraging finding indicates that those with a history of mistreatment may have just as many core social network members to turn to as those without mistreatment experience. Interventions can engage core social network members to maintain supportive relationships with mistreatment victims.

Other analyses of the structure of the two mistreated groups’ social networks showed distinct patterns warranting tailored recommendations. Those reporting polyvictimization had fewer kin in their core social networks and had less dense core social networks. Again, this pattern of disconnectedness from family members suggests that we cannot assume all older adults can rely on relatives as primary sources of support. However, increasing network density by fostering communication across older adults’ existing core social network members could improve well-being for polyvictimization victims, especially given that their core social networks were similar in size to those in the mistreatment unlikely class.

The financial mistreatment class also had a lower percentage of kin in their core social networks. They specifically had a lower percentage of children in their core social networks despite being more likely to live with a child. Perhaps older adults who lived with children had a higher risk of experiencing financial mistreatment because coresident children who may have provided care to these older adults felt entitled to, and may have had access to, older adults’ financial resources (DeLiema et al., 2018; Gibson & Qualls, 2012). Therefore, supporting financial mistreatment victims may involve minimizing coresidence with children or providing outside assistance with management of finances.

One limitation of this study was that NSHAP’s broadly phrased lifetime mistreatment questions could overidentify mistreatment victims. Yet, robustness checks using stricter criteria for detecting victimization produced qualitatively similar results. In descriptive analyses, both methods agreed on the most common mistreatment behaviors and perpetrators. The latent class analysis produced the same three classes of mistreatment victims regardless of whether we accounted for mistreatment severity. Finally, of the 22 estimates of interest produced in the 11 regression models of victims’ social contexts, only two results that were statistically significant in the main models did not remain so, even at the less stringent level of p < .10, in the robustness checks. These deviations were likely due to lower statistical power in the secondary models. Thus, we trust our findings regarding mistreatment history and current social network characteristics.

Future work could address the study’s other limitations. Researchers could consider whether victimized older adults may have also mistreated their children or relatives to better contextualize the social dynamics of elder mistreatment in further developing interventions (Knobloch, Nichols, & Martindale-Adams, 2019; Pickering, Yefimova, Maxwel, Puga, & Sullivan, 2019). Such interventions may include alleviating the distress that affects those helping elder mistreatment victims (Breckman et al., 2017). Future researchers could also examine how caregiving shapes the social context of elder mistreatment to inform the design of further social support interventions. Despite these limitations, this study offers information on mistreatment victims’ social contexts that can guide the development of social support interventions.

Funding

This work was supported by the National Institute on Aging and the National Institutes of Health (R01AG043538, R01AG048511, and R37AG030481). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflict of Interest

None reported.